BRD4 binds to active cranial neural crest enhancers to regulate RUNX2 activity during osteoblast differentiation

Abstract

Cornelia de Lange syndrome (CdLS) is a congenital disorder featuring facial dysmorphism, postnatal growth deficits, cognitive disability and upper limb abnormalities. CdLS is genetically heterogeneous, with cases arising from mutation of BRD4, a bromodomain protein that binds and reads acetylated histones. In this study, we have modeled CdLS facial pathology through mouse neural crest cell (NCC)-specific mutation of BRD4 to characterize cellular and molecular function in craniofacial development. Mice with BRD4 NCC loss of function died at birth with severe facial hypoplasia, cleft palate, mid-facial clefting and exencephaly. Following migration, BRD4 mutant NCCs initiated RUNX2 expression for differentiation to osteoblast lineages but failed to induce downstream RUNX2 targets required for lineage commitment. BRD4 bound to active enhancers to regulate expression of osteogenic transcription factors and extracellular matrix components integral for bone formation. RUNX2 physically interacts with a C-terminal domain in the long isoform of BRD4 and can co-occupy osteogenic enhancers. This BRD4 association is required for RUNX2 recruitment and appropriate osteoblast differentiation. We conclude that BRD4 controls facial bone development through osteoblast enhancer regulation of the RUNX2 transcriptional program.

Keywords: BRD4; Cornelia de Lange syndrome; Craniofacial; Histone acetylation reader; Mouse; Neural crest.

Fig. 1.

BRD4 NCC loss of function produces severe craniofacial phenotypes. (A) Schematic of BRD4 protein with locations of bromodomains (Bromo), extraterminal domain (ET), exon 5 coding sequence that is removed by Cre/LoxP and amino acid Y432 where point mutation produces Cornelia de Lange syndrome. (B-I) Lateral and side view images of E18.5 wild-type (WT) (B,C), Brd4^cW1KO (D,E), more severe Brd4^cS10KO (F,G) and less severe Brd4^cS10KO (H,I) embryos highlighting exencephaly (red arrows), mid-facial clefting (blue arrows) and open eye phenotypes (yellow arrows). All Brd4^cW1KO and Brd4^cS10KO embryos demonstrate severe anterior facial hypoplasia with smaller frontal, nasal and mandible regions. (J) Summary of Brd4^cW1KO and Brd4^cS10KO phenotypic frequencies (N=5 and N=24, respectively). (K-P) Alizarin Red and Alcian Blue stain of bone and cartilage wholemount images of E18.5 WT and Brd4^cS10KO embryos with ventral wholemount view (K,L), dissected mandible (M,N) and ventral cranial base view (O,P) highlighting micrognathia (white arrows), basisphenoid bone (aqua arrows) and presphenoid bone formation (black arrow). Scale bars: 5 mm (K,L); 2 mm (M,N); 3 mm (O,P).

Fig. 2.

BRD4 mutant mandibular cNCCs fail to properly differentiate to osteoblast lineages. (A-D) Brightfield and Rosa^Tomato reporter fluorescence wholemount imaging of E11.5 Brd4^cS10Het control and Brd4^cS10KO embryos had similar Rosa^Tomato+ cNCC domains. Dashed white line in A depicts sectioning region for immunofluorescence at E11.5 in E-J. (E-J) Immunofluorescence within coronal sections of the E11.5 wild-type (WT) and Brd4^cS10KO first branchial arch for activated cleaved Caspase-3 (Casp.; E,F), bromodeoxyuridine (BrdU; G,H) incorporation and phosphorylated histone H3 serine 10 (pH3S10; I,J) along with DAPI (blue) demonstrated normal proliferation and lack of apoptosis in Brd4^cS10KO embryos. (K-N) Brightfield and Rosa^Tomato reporter fluorescence at E13.5 illustrated loss of cNCC domains in Brd4^cS10KO embryos relative to Brd4^cS10Het controls. Dashed white line in K depicts sectioning region for immunofluorescence at E13.5 in O-T. (O-T) Immunofluorescence within coronal sections of the E13.5 WT and Brd4^cS10KO developing mandible for BRD4 (O,P), RUNX2 with type II collagen (COL2; Q,R), and RUNX2 with Osterix (OSX; S,T). Brd4^cS10KO RUNX2⁺ pre-osteoblasts fail to induce Osterix expression. All immunofluorescence images are overlaid with DAPI nuclear stain (blue). Scale bars: 1 mm (A-D,K-N); 200 µm (E-J,O-T).

Fig. 3.

Loss of BRD4 disrupts in vitro cNCCs osteoblast differentiation. (A-D) BRD4 immunofluorescence in wild-type (WT), hypomorphic (Brd4^hypo) with trans-heterozygous frameshift (fs) mutations in exon3, or knockout (Brd4^KO1 or Brd4^KO2) cNCC cell lines with trans-heterozygous fs mutations in exons 3 and 5. DAPI nuclear stain shown in blue. (E) Western blot of BRD4 hypomorphic and knockout cNCC lines demonstrated loss of BRD4 relative to nucleolin (NLN) loading control. (F) Western blot of osteochondral transcription factors SOX9 and RUNX2 are unaltered in BRD4 knockout cNCCs relative to NLN loading control. (G) Flow cytometry histogram of Cell Trace Far Red (CTFR) tracking dye demonstrated gradual dilution as WT cNCCs proliferate across 5 days of growth. (H,I) Cells labeled with similar levels of CTFR dye at onset (H) revealed slightly slower proliferation rates for BRD4 hypomorphic and BRD4 knockout cNCC lines compared with WT at day (D)3 of growth (I). (J-L) At D7 of osteogenic differentiation, compared with WT, Brd4^KO1 and Brd4^KO2 lines lack detectable alkaline phosphatase activity (Alk. Phos.). (M-O) At D10 of osteogenic differentiation, WT cNCCs exhibited robust alkaline phosphatase activity that was diminished in Brd4^KO1 and lost in Brd4^KO2. (P-S) At D7 of differentiation, WT first branchial arch primary cNCCs (WT BA D7) exhibit similar alkaline phosphatase activity as O9-1 cell culture (WT cNCC D7); however, Brd4^cS10KO primary cNCCs (BA D7) fail to differentiate (S). Scale bars: 10 µm (A-D); 2 mm (J-O); 1 mm (P-S).

Fig. 4.

BRD4 binds to proximal active enhancers to regulate osteogenic transcription. (A) Venn diagram plots of significantly altered (logFC≥1 or≤−1) expressed genes (WT RPKM≥1) from Brd4^KO1 or Brd4^KO2 compared with wild-type (WT) in day (D)0 undifferentiated cNCCs or at D3 and D6 of osteogenic differentiation. Both Brd4^KO1 and Brd4^KO2 lines demonstrated overlap of upregulated and downregulated genes. (B) UCSC genome browser tracks of BRD4 binding in ESCs (blue), BRD4 binding in D0 undifferentiated cNCCs or at D3 and D6 of osteogenic differentiation in WT (purple) or Brd4^KO2 (KO, black) cells. Also illustrated are enhancer histone modifications including H3K27ac (pink) in WT D0 undifferentiated cNCCs or at D3 and D6 of osteogenic differentiation and H3K4me2 accumulation (green) in WT D0 undifferentiated cNCCs. RUNX2 binding (red) in WT D0 cNCC, at D3 of osteogenic differentiation and MC3T3 pre-osteoblasts highlighted RUNX2 enrichment [osterix (Sp7) and Adamts4] at BRD4 sites, reduction in Brd4^KO2 (D0 KO) and absence in ESC controls (black). H3K27me3 accumulation (red) in WT D0 undifferentiated cNCCs illustrated repressive chromatin regions. Gene loci of interest are osterix, Col1a1, Adamts4, Malat1, and Hoxb gene loci. BRD4 is bound (red arrows) to active enhancers of target genes featuring high levels of H3K27ac, H3K4me2 and RUNX2 binding.

Fig. 5.

BRD4 directly regulates transcription of factors that are crucial for osteoblast differentiation. (A) MSigDB canonical pathways that overlap with the top 500 day (D)3 and D6 osteogenic BRD4 targets (Tables S2,S3, sheet 2). (B) MSigDB human phenotype ontogeny pathways that overlap with the top 500 D3 and D6 osteogenic BRD4 targets (Tables S2,S3, sheet 2). (C) Key for volcano plots in D,E. Significantly upregulated genes are colored green. Significantly downregulated genes are colored red if they do not change across osteogenic differentiation, blue if they increase in wild-type (WT) expression at D3 of osteogenic differentiation, purple if they increase in WT expression at D6 of osteogenic differentiation. BRD4 directly bound, downregulated targets have filled circles. (D) Volcano plot of log₂ fold change versus −log₁₀ false discovery rate (FDR) comparing D3 osteogenic WT expression with Brd4^KO2. Only gene sets in common with Brd4^KO1 are color coded. (E) Volcano plot of log₂ fold change versus −log₁₀ false discovery rate comparing D6 osteogenic WT expression with Brd4^KO2. Only gene sets in common with Brd4^KO1 are color coded. In D,E, BRD4 binds directly to regulate expression of large sets of genes that are crucial for osteoblast differentiation. (F) Numbers of BRD4 bound enhancers versus promoters for target genes (Tables S1-S3, sheet 2). (G) Comparison of super-enhancer frequency for BRD4 bound enhancers compared with annotated enhancers lacking BRD4 binding. (H) BRD4 downregulated direct targets at D3 and D6 of osteogenic differentiation (Tables S2,S3, sheet 2) were compared for overlap with downregulated genes in E13.5 Brd4^cS10KO embryonic cNCCs (Table S4) and charted as percentage overlap. (I-L) Immunofluorescence on coronal sections of the E13.5 WT and Brd4^cS10KO developing mandible for RUNX2 with type I collagen (COL1A1; I,J), or FGFR2 (K,L) along with DAPI (blue). Brd4^cS10KO RUNX2⁺ pre-osteoblasts fail to induce COL1A1 and FGFR2 expression. Scale bars: 200 μm.

Fig. 6.

BRD4 associates with RUNX2 to regulate osteoblast differentiation. (A) Enrichment of DNA transcription factor binding motifs were analyzed at day (D)3 or D6 BRD4 bound target osteogenic enhancers (Tables S2,S3, sheet 4) using the HOMER findMotifsGenome.pl program. RUNX2 motifs were enriched at both time points compared with BRD4 unbound enhancers. (B) D3 RUNX2 CUT&RUN relative read density was plotted at BRD4 stem cell enhancers (Table S1, sheet 4), D3 BRD4 osteogenic enhancers (Table S2, sheet 4), D6 osteogenic enhancers (Table S3, sheet 4) or enhancers not bound by BRD4. RUNX2 demonstrated enrichment at BRD4 osteogenic and stem cell enhancers. (C) Profile of counts per million mapped reads (CPM) normalized RUNX2 enrichment in wild-type (WT; blue) or Brd4^KO2 cells (red) at D0 stem cell enhancers (left) or D3 BRD4 osteogenic enhancers (right). CUT&RUN for RUNX2 in ESCs (black) served as a negative control due to lack of expression in these stem cells. (D) Structure of BRD4 protein with reference to short or long isoforms. (E) Flag-tagged BRD4 constructs were co-transfected into HEK293T with HA-tagged RUNX2 followed by immunoprecipitation (IP) on Flag antibody conjugated beads. The C-terminus of BRD4 encoded by the long isoform is responsible for protein IP of RUNX2. (F) Lentiviral transduction of the human BRD4 long isoform was capable of restoring Brd4^KO2 osteoblast differentiation after 10 days of differentiation, whereas BRD4 short isoform was not capable of supporting differentiation. Images depict alkaline phosphatase (Alk. Phos.) activity on substrate colorimetric reaction. (G) Lentiviral transduction and overexpression of RUNX2 was capable of restoring Brd4^KO2 osteoblast differentiation after 10 days of differentiation. Scale bars: 2 mm.

Fig. 7.

Model of BRD4 function in Cornelia de Lange syndrome craniofacial pathogenesis. Our results indicate that BRD4 binds to enhancers to induce transcription of osteogenic genes and proper cNCC osteoblast differentiation. BRD4 is required for efficient RUNX2 recruitment to drive appropriate expression of the RUNX2 transcriptional program during osteogenic differentiation. Although BRD4 also associates with the NIPBL cohesin loading protein, the predominantly mutated factor in CdLS, the role of this association in osteogenic enhancer activity and craniofacial development is unknown. Figure created with BioRender.